[llvm-pdbutil] Fix register enum field dumping/parsing

[nikitalita]

This fixes a bug where parsing PDBs with usages of register enums were asserting.

The main problem is that printing out the code view register enums are taken care of here:
https://github.com/nikitalita/llvm-project/blob/e4888a92402f53000a3a5e79d3792c034fc2f343/llvm/lib/ObjectYAML/CodeViewYAMLSymbols.cpp#L152

Which requires a COFF::header in the IO context for the machine type, which we didn't have when dumping a pdb or parsing a yaml file. So, we make a fake one with the machine type.

[llvmbot]

@llvm/pr-subscribers-debuginfo

Author: None (nikitalita)

Changes

The main problem that we were encountering is that printing out the code view register enums are taken care of here:
https://github.com/nikitalita/llvm-project/blob/e4888a92402f53000a3a5e79d3792c034fc2f343/llvm/lib/ObjectYAML/CodeViewYAMLSymbols.cpp#L152

Which requires a COFF::header in the IO context for the machine type, which we didn't have when dumping a pdb or parsing a yaml file. So, we make a fake one with the machine type.

---

Full diff: https://github.com/llvm/llvm-project/pull/82299.diff

3 Files Affected:

• (modified) llvm/tools/llvm-pdbutil/PdbYaml.cpp (+7)
• (modified) llvm/tools/llvm-pdbutil/PdbYaml.h (+2)
• (modified) llvm/tools/llvm-pdbutil/YAMLOutputStyle.cpp (+9)

diff --git a/llvm/tools/llvm-pdbutil/PdbYaml.cpp b/llvm/tools/llvm-pdbutil/PdbYaml.cpp
index a26241967b5add..fac1d89321610b 100644
--- a/llvm/tools/llvm-pdbutil/PdbYaml.cpp
+++ b/llvm/tools/llvm-pdbutil/PdbYaml.cpp
@@ -155,6 +155,13 @@ void MappingTraits<PdbDbiStream>::mapping(IO &IO, PdbDbiStream &Obj) {
   IO.mapOptional("PdbDllRbld", Obj.PdbDllRbld, uint16_t(0U));
   IO.mapOptional("Flags", Obj.Flags, uint16_t(1U));
   IO.mapOptional("MachineType", Obj.MachineType, PDB_Machine::x86);
+  // This is a workaround for IO not having document context with the
+  // machine type. The machine type is needed to properly parse Register enums
+  // in the PDB.
+  if (!IO.getContext()) {
+    Obj.FakeHeader.Machine = static_cast<uint16_t>(Obj.MachineType);
+    IO.setContext(&Obj.FakeHeader);
+  }
   IO.mapOptional("Modules", Obj.ModInfos);
 }
 
diff --git a/llvm/tools/llvm-pdbutil/PdbYaml.h b/llvm/tools/llvm-pdbutil/PdbYaml.h
index 4382e91e209737..b0c0f6a00c499b 100644
--- a/llvm/tools/llvm-pdbutil/PdbYaml.h
+++ b/llvm/tools/llvm-pdbutil/PdbYaml.h
@@ -11,6 +11,7 @@
 
 #include "OutputStyle.h"
 
+#include "llvm/BinaryFormat/COFF.h"
 #include "llvm/DebugInfo/CodeView/SymbolRecord.h"
 #include "llvm/DebugInfo/CodeView/TypeRecord.h"
 #include "llvm/DebugInfo/MSF/MSFCommon.h"
@@ -80,6 +81,7 @@ struct PdbDbiStream {
   PDB_Machine MachineType = PDB_Machine::x86;
 
   std::vector<PdbDbiModuleInfo> ModInfos;
+  COFF::header FakeHeader;
 };
 
 struct PdbTpiStream {
diff --git a/llvm/tools/llvm-pdbutil/YAMLOutputStyle.cpp b/llvm/tools/llvm-pdbutil/YAMLOutputStyle.cpp
index 80b76657facc7c..ecb4c2175e49a3 100644
--- a/llvm/tools/llvm-pdbutil/YAMLOutputStyle.cpp
+++ b/llvm/tools/llvm-pdbutil/YAMLOutputStyle.cpp
@@ -11,6 +11,7 @@
 #include "PdbYaml.h"
 #include "llvm-pdbutil.h"
 
+#include "llvm/BinaryFormat/COFF.h"
 #include "llvm/DebugInfo/CodeView/DebugChecksumsSubsection.h"
 #include "llvm/DebugInfo/CodeView/DebugSubsection.h"
 #include "llvm/DebugInfo/CodeView/DebugUnknownSubsection.h"
@@ -73,7 +74,15 @@ Error YAMLOutputStyle::dump() {
   if (auto EC = dumpPublics())
     return EC;
 
+  // Fake Coff header for dumping register enumerations.
+  COFF::header Header;
+  auto MachineType =
+      Obj.DbiStream ? Obj.DbiStream->MachineType : PDB_Machine::Unknown;
+  Header.Machine = static_cast<uint16_t>(MachineType);
+  Out.setContext(&Header);
   flush();
+  Out.setContext(nullptr);
+
   return Error::success();
 }
 

[github-actions]

⚠️ We detected that you are using a GitHub private e-mail address to contribute to the repo.
Please turn off Keep my email addresses private setting in your account.
See LLVM Discourse for more information.

[nikitalita]

fixes #58916

[tru]

I think we need some kind of test to cover this case.

[nikitalita]

Never mind, I found one and I have added unit tests

[nikitalita]

The tests are passing now.

[nikitalita]

@tru can you please take a look?

[tru]

I don't know enough about the PDB spec to say if this is the best way to solve it. The code looks fine enough.

Maybe @mstorsjo @aganea or @rnk could help

[rnk]

Thanks, looks good.

[nikitalita]

@rnk I have rebased on master, can you push the button?